Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/XxYouDeaDPunKxX/canon-boundary-guard-codex/llms.txt

Use this file to discover all available pages before exploring further.

Canon Boundary Guard gives Codex a structured classification frame for information provenance. When Codex works on a repository it reads more than just files — it also sees conversation history, runtime context, operator instructions, and model assumptions. Those sources do not carry the same authority. Canon Boundary Guard makes the boundaries between them visible so the right material becomes persistent project content and nothing else does.

Introduction

Understand why provenance boundaries matter and what Canon Boundary Guard does

Installation

Add the plugin to Codex in three steps using the marketplace command

Quickstart

Activate the frame, verify it’s working, and run your first guarded session

Provenance Layers

Learn the six source classes — L0 through L3 — and when each applies

​
Why Canon Boundary Guard

Codex often works with several kinds of information at the same time:
  • Project files — source code, schemas, lockfiles, git state
  • Conversation context — messages from the current chat thread
  • Operator instructions — directions given to Codex about how to behave
  • Runtime context — AGENTS.md guidance loaded at session start
  • Model assumptions — unverified conventions and best-practice priors
Without a classification frame, these sources blur together. A working hypothesis from the chat might silently become a project file. An instruction to the agent might get written into repository content. A model assumption might override verified evidence. Canon Boundary Guard makes those boundaries explicit and keeps them visible throughout the session — during reading, reasoning, planning, and ultimately writing.

​
Core Concepts at a Glance

Provenance Layers

Six source classes from verified evidence (L0) to model priors (L3)

AGENTS.md Authority

How Canon Boundary Guard classifies runtime instruction-chain preludes

Inline Tagging

Mark non-L0 material inline so provenance stays visible in generated output

Dossier Modes

Three modes for tracking edit evidence: mechanical, semantic, and promotion

​
Guides

Using the Skill

Activate Canon Boundary Guard and keep it running for a full session

Hook Setup

Wire the PreToolUse hook so the frame re-surfaces before every write

Decontamination

Identify and remove conversation residue, agent-control language, and model priors

Session Workflow

End-to-end workflow from session start through compaction and write decisions
Canon Boundary Guard is licensed under CC BY-SA 4.0. It was developed with AI assistance under human direction.

Build docs developers (and LLMs) love

Get started for freeTalk to us